HAL Id: hal-00250188
https://hal.archives-ouvertes.fr/hal-00250188
Submitted on 11 Feb 2008
HAL
is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire
HAL, estdestinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Performance Analysis of IEEE 802.15.4 Contention Free Period through Real-Time Industrial Maintenance
Applications
Nicolas Salles, Nicolas Krommenacker
To cite this version:
Nicolas Salles, Nicolas Krommenacker. Performance Analysis of IEEE 802.15.4 Contention Free Pe-
riod through Real-Time Industrial Maintenance Applications. 22nd IAR annual meeting, Nov 2007,
Grenoble, France. pp.CDROM. �hal-00250188�
PERFORMANCE ANALYSIS OF IEEE 802.15.4 CONTENTION FREE PERIOD THROUGH REAL-TIME INDUSTRIAL MAINTENANCE
APPLICATIONS
Nicolas Salles∗Nicolas Krommenacker∗
∗Centre de Recherche en Automatique de Nancy Nancy-Université, CNRS, (CRAN – UMR 7039)
Faculté des Sciences et Techniques, BP 239, F-54 506 Vandœuvre-lès-Nancy Cedex,France
Abstract: Nowadays, a growing number of societies are looking for industrial wireless network solutions. However, industrial communications induce real-time constraints and those networks should at least fulfil those requirements. In this paper, we study IEEE 802.15.4 performances in order to support real-time traffic.
This analysis brings out some limitations of that standard concerning cycle duration and amount of messages per cycle.
Keywords: IEEE 802.15.4, industrial wireless, performance analysis, real-time.
1. INTRODUCTION
Over the past few years, intensive wireless net- works developments have been done and their usage for industrial cases appear more and more possible. In fact, wireless networks provide notice- able advantages in terms of mobility and costs reduction. Nevertheless, transmission errors may occur and are not negligible. So wireless commu- nications are not advisable for control-command systems forasmuch as hard real-time systems Nonetheless, they may be used for subsystems like monitoring and e-maintenance applications (Ramamurthy et al., 2007) where a loss of a few information will not be a major inconvenience (i.e.
an operator provided with an handled device).
IEEE 802.15.4(IEEE Computer Society, 2006), one of the most recent wireless standard, is quite interesting due to its “ultra-low complexity, low- cost and extremely low-power wireless connectiv- ity” (Gutiérrez et al., 2003). That protocol pro- vides two different access methods to the medium.
The mandatory one is CSMA/CA (Carrier Sense Multiple Access / Collision Avoidance) which is not deterministic by design (possible infinite re-
emissions). This mechanism may provide a de- terministic medium access by the use of an ap- plication layer based on a master/slave protocol.
The second medium access method is optional and achieves a mechanism of time reservation called GTS (Guaranteed Time Slots) These time allo- cations may allow periodic unidirectional trans- missions within a bounded time delay between a device and its coordinator. Such reservations are bounded to seven GTS’s in a duty cycle. So their use are restricted to periodical pooling applica- tions by a coordinator device of a limited number of equipments.
In this paper, we present a performance analysis between the two medium access mechanism pro- vided by IEEE 802.15.4. We will present there- fore some advantages and inconveniences for both methods.
This paper is organised as follow. In Section 2 we overview the IEEE 802.15.4 specification. Sec- tion 3 comments on network configuration. Sec- tion 4, through two different network topologies, focuses on performance analysis of GTS mech- anism based on an industrial case. Then it is
compared to a CSMA/CA based medium access in Section 5. Section 6 relates our contribution with previous works on performance analysis of industrial wireless communications. Finally, this paper concludes in Section 7.
2. IEEE 802.15.4 OVERVIEW 2.1 Network design
IEEE 802.15.4 defines two different device types; a full-function device (FFD) and a reduced-function device (RFD). So an FFD is able to communicate with either type of devices, whereas an RFD can only discuss with an FFD.
FFD can operate in three modes serving as a personal area network (PAN) coordinator, a co- ordinator, or a device. Differences between PAN coordinator and coordinator lie on their range of action. A PAN coordinator controls every coordi- nator in its PAN whereas a coordinator takes only care of directly connected devices. Later, we won’t make any differences as our examples consider only one coordinator (PAN coordinator).
An RFD is intended for simple applications, such as a sensor, it does not need to send large amount of data and may only be associated with an FFD at a time. So, RFD may use low resources and memory capacity and may be cheaper than FFD.
On a plant floor, we consider most of sensors are RFDs.
IEEE 802.15.4 standard defines two types of topologies (fig. 1).
Full Function Device PAN
coordinator
(a) star
Reduced Function Device PAN coordinator
(b) peer to peer
Figure 1. IEEE 802.15.4 network topologies Star topology restricts every device to commu- nicate only with their PAN coordinator whereas peer-to-peer one allows communication between devices if they are able to. Star topology may result in use of a mains powered coordinator and battery powered nodes. This paper focuses on star topology through a study case based on an European Integrated Project(DYNAMITE — Dynamic Decisions in Maintenance, 2005–2009).
2.2 Medium access control
IEEE 802.15.4 medium access may be synchro- nised or not through a beacon frame. So the stan-
dard specifies the use of a periodic beacon frame to synchronise devices in beacon-enabled mode whereas it is also possible to communicate without any synchronisation in non-beacon-enabled mode.
Both modes provides medium access within a Contention Access Period (CAP). Beacon-enabled mode provides in addition a Contention Free Period (CFP). Therefore, we focus on beacon- enabled mode.
Between beacons, access to the medium is part in two (see fig. 2). The first part, called “super- frame”, represents time allocated by the coordi- nator for communication whereas the second one corresponds to network inactivity for every node related to that coordinator.
beacon beacon beacon
active period
beacon interval superframe inactive
period active
period inactive period
Figure 2. Beacon-enabled access mode
Beacon interval (BI) and superframe (SD) dura- tions are defined by the coordinator through the two parameters Beacon Order (BO) and Super- frame Order (SO) as follow:
BI=aBSFD×2BO (1) SD=aBSFD×2SO (2)
0≤SO≤BO≤14 (3)
Period of network inactivity purposes some power- saving sought by IEEE 802.15.4 standard. In our study, we consider the shortest inactive period, as we are looking for communication performances without considering energy consumption contrary to (Koubâa et al., 2006a).
As the inactive period may not exist (SO = BO), we only consider the active period, super- frame, composed of a beacon frame, Contention Access Period (CAP) and a Contention Free Pe- riod (CFP) (see fig. 3). Then we consider cycle duration is equivalent to superframe duration.
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
beacon beacon
superframe
CAP CFP
GTS GTS
Figure 3. Superframe composition
2.3 Contention Free Period
By definition, a superframe is divided into 16 equals slots which may be allocated to a device by the coordinator for dedicate communication.
Those kind of reservations are called Guaran- teed Time Slots (GTS) and may include an in- teger number of these slots. Any device can ask for such a reservation as long as there are no more than 7 GTS’s and CAP duration is at least aMinCAPLength long.
2.4 Superframe duration
Superframe duration SD is defined by the stan- dard by (2) and (3) where aBSF D = aBSD× aNSFD. aBSD and aNSFD parameters are de- fined by the standard.
So, it is possible to determine, every superframe duration depending on the network parameters set by the coordinator (channel/frequency, mod- ulation and superframe order). we figure out that most efficient network parameters implies the use of a O-QPSK modulation in association with a frequency in the range either 902–928 MHz or 2 400–2 483.5 MHz. The use of such a configura- tion results in superframe duration included be- tween 15.36ms and 251.6 s. Due to the fact the first range of frequencies is optional for standard specification while the second one is mandatory, we only take care of the use of band 2 400–
2 483.5 MHz in that paper. So the remaining parameter the coordinator has to define, is the superframe order.
3. NETWORK CONFIGURATION In order to evaluate a superframe duration, we have to know how a network is configured. So we need to specify the value of the undefined param- eter, superframe orderSO. More the superframe order is low, more the superframe duration is low and so more the cycle duration is accurate for real-time communications. Choice of superframe order must be as low as possible also it respects constraints concerning standard specification and usage of the network. In the next two subsections, we focus on those two constraints.
3.1 Specification constraints
CFP usage implies the respect of the transmis- sion of the beacon frame as well as the minimum CAP duration,aMinCAPLength. Then in order to evaluate the impact of a superframe order value, we must consider the mandatory slots Mslots re- quired by these two transmissions. CAP length is defined by the coordinator and so it is consid- ered as the minimum required by the standard aMinCAPLength. In contrary, beacon frame size, which is application dependent, is considered as its maximum sizeBeaconmax= 133bytesheaders
included. Thereby, we determine (4),(5) the maxi- mum CFP length in terms of remaining slotsRslots
depending on the superframe order (see fig. 4).
Mslots=
aMinCAPLength+Beaconmax
aBaseSlotDuration×2SO
(4)
Rslots= 16−Mslots (5)
0 4 8 12 16
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Superframe orderSO
Rslots b b b b b b b b b b b b b b b
2400–2483,5 MHz O-QPSK
Figure 4. Maximum available slots for CFP So, choice of a superframe can be checked over the minimum required by the specification. For example, if CFP has to be 5 slots long, it may be impossible to satisfy all requirements with SO = 0. In such a case, superframe order must be chosen at least equal to 1.
3.2 Usage constraints
In order to evaluate the choice of a superframe or- der, we must define how many slots are needed for CFP. In that part, we deal with CFP slots require- ments depending on data transmission. We con- sider a unidirectional transmission between two devices ofDatabytes in one GTS. That amount of data is considered at the application layer and will be encapsulated by the two IEEE 802.15.4 layers in order to form a network packet. For a com- munication in a GTS, we consider PHYheaders= 6 bytes and MACheaders = 23 bytes. MACheaders
is defined considering the maximum value for aux- iliary security field as we can’t control the use of such a parameter in industrial environments.
For 1 byte of payload data, we send at least 30 bytes over the network including 29 bytes corre- sponding to encapsulation. In addition, we should also take care of transmission scheme specified by the standard and notably of interframe spaces (IFS) which should be part of GTS. Figure 5 explains that mechanism which should follow ev- ery packets send over the network. It depends size of transmitted packets and acknowledgement configuration.
Short frames correspond to packets which length is lower or equal to 24 bytes. Due to the en- capsulation considered in that paper, we doesn’t develop what concerns short frames and only con- sider transmission of long frames in either case of acknowledgement. So, we define in (6), the total amount of bytes Datatransmitted transmitted over the network considering minimum fragmentation.
Long frame
LI F S
Short frame SI F S (a) unacknowledged Long frame
tack
ACK LI F S
Short frame tack
ACK SI F S (b) acknowledged
Figure 5. IEEE 802.15.4 interframes spaces
Datatransmitted=Data+
Data M SDUmax
×LH (6)
LH=
(Headers unacknowledged
Headers+tACK+ACK acknowledged
Headers=MACheaders+PHYheaders+LIF S Then we look for the slots needed by transmission of data payload ofDatabytes in a GTS. The num- berN of slots is provided by (7) and represented on figure 6.
Nslots=
Datatransmitted
aBaseSlotDuration×2SO
(7) Slots needed for a transmission can be evaluate and bring face to face with specification require- ments (8).
X
i
Nslotsi ≤Rslots (8)
As an example, if we want to send 60 unac- knowledged bytes on the network, we may select SO = 0, Nslots = 4 slots with (7) which satisfy the maximum Rslots (5). But in the case data need to be acknowledged superframe order set to 0 induces Nslots = 5 slots and it contradicts the maximumRslots = 4 slots obtained by (5). So in that case, a solution may beSO= 1 which induces (Nslots= 3 slots)≤(Rslots = 4 slots).
4. INDUSTRIAL CASE
Now we are able to set network parameters in order to satisfy the minimum requirements of such a network, we look for available real-time abilities based on an industrial case. Our study case is based on maintenance of industrial plants.
On a plant floor, we want to provide to main- tenance teams with direct access to machines’
sensors in order to know in real-time how they progress. Maintenance people carries a mobile de- vice (PDA) which is able to communicate with plant’s sensors using wireless networks.
That paper studies such capacities regarding the use of IEEE 802.15.4 standard.
4.1 Simple infrastructure
As a first evaluation, we study a network based on a star topology (fig. 1(a)) where PAN coordinator is a PDA and other devices are only RFDs.
In such a configuration every sensor in PDA’s range of action are able to send data for main- tenance purpose to the PDA after a small time of auto-configuration (negotiation, association, . . . ).
Our interest is to specify what would be the minimum cycle duration to ensure communication of every devices with the PDA in relation with network’s requirements.
For that study, we consider a network set up with five industrial sensors transmitting and/or receiving data as described in table 1.
Table 1. Data exchanges
Device Data send Data received Cycle
no1 1 byte - 40 ms
no2 8 bytes - 40 ms
no3 4 bytes - 60 ms
no4 8 bytes 4 bytes 60 ms
no5 16 bytes - 100 ms
Even if sensors doesn’t require the same period of cycle, we consider a cycle duration which allows any device to transmit during any superframe.
Then the defined set of transmissions induce 6 GTS’s; emission and reception implies for node 4 the use of two different GTS’s.
Such a configuration lead us to deduce the best network configuration corresponding to that us- age. For that purpose, we compute Ni for every node and we evaluate CFPslots = ΣiNi. Evalua- tion of superframe order is given table 2 and 3.
Table 2. Superframe order evaluation, unacknowledgement
SO 0 1 2 3
N1 2 1 1 1
N2 2 1 1 1
N3 2 1 1 1
N4 2 1 1 1
N4′ 2 1 1 1
N5 3 2 1 1
CFPslots 13 7 6 6
CFPmax 4 10 13 14 SD(ms) 15.36 30.72 61.44 122.88
For that specific case, communications will be insured with the choice of SO = 2 with ac- knowledgement whereas SO = 1 is sufficient for unacknowledged communications. SO = 1 corre- sponds to cycle duration equals 30.72 ms which satisfies all requirements (Cycle≥40ms) whereas superframe duration of 61.44 ms doesn’t. So we are able to satisfy sensors’ communications in case we set up a network with SO = 1 without acknowledgement.
0 1 2 3 4 5 6 7 8 9 10
0 20 40 60 80 100 120 140 160 180
Data to be transmittedData (bytes)
SlotsusageN
1 packet 2 packets
SO= 0
SO= 1 SO= 2 2400–2483,5 MHz
O-QPSK
(a) unacknowledged
0 1 2 3 4 5 6 7 8 9 10 11 12
0 2 4 6 8 10 12 14 16 18
Data to be transmittedData(bytes)
SlotsusageN
1 packet 2 packets
SO= 0
SO= 1
SO= 2 2400–2483,5 MHz
O-QPSK
(b) acknowledged
Figure 6. Slots usage in relation with data transmission Table 3. Superframe order evaluation,
acknowledgement
SO 0 1 2 3
N1 3 2 1 1
N2 3 2 1 1
N3 3 2 1 1
N4 3 2 1 1
N4′ 3 2 1 1
N5 4 2 1 1
CFPslots 19 12 6 6
CFPmax 4 10 13 14 SD(ms) 15.36 30.72 61.44 122.88
That study considers only sporadic communica- tions from sensors to an unique receptor. But in industrial cases, data may be useful for more than one recipient and that is why we should take into account cases with a more permanent structure.
4.2 Permanent infrastructure
Introduction of a fixed coordinator on a plant floor leads to a more permanent infrastructure (see fig. 7). In fact the presence of a coordinator allows communication from the sensors to the coordinator even if there is no maintenance PDA in the area. Moreover, it induces a reduction in time needed for association.
6 1
2 4 3
PAN 5 coordinator
PDA
Figure 7. Permanent infrastructure
In terms of performances we consider the example in addition with the communication between coor- dinator and PDA. This data transmission includes all the previously received frames by the coordina- tor from sensors. Therefore, this communication includes the same number of frames and we are unable to use (7) with Data = P5
i=1Datai. In order to calculate the number of slots required
for the transmission to PDA, we need to compute Datatransmitted6=P5
i=1Datatransmittedi
In our case, all frames are lower thanM P DUmax (127 bytes), so we reduce the expression to Datatransmitted6 = P5
i=1Datai + 5×LH. Then we deducts the superframe order corresponding to that network (tables 4 and 5).
Table 4. Superframe order evaluation, unacknowledgement
SO 0 1 2 3
ΣNslots 11 11 8 7
Rslots 4 10 13 14
SD(ms) 15.36 30.72 61.44 122.88
Table 5. Superframe order evaluation, acknowledgement
SO 0 1 2 3
ΣNslots 30 18 9 8
Rslots 4 10 13 14
SD(ms) 15.36 30.72 61.44 122.88
In that case, superframe duration is 61.44 ms in either case and we can’t satisfy the requirements of all sensors. Moreover we must take into consid- eration that, with that kind of infrastructure, data arrived at PDA after two superframes and so we should consider cycle duration equals 122.88ms.
5. METHODS COMPARISON
GTS access method provides a simple polling mechanism for IEEE 802.15.4 devices. It may also be possible to implement a polling protocol through CSMA/CA.
In order to compare use of a CFP with CSMA/CA mechanisms, we consider such a polling protocol requiring an overhead of 1 byte. Coordinator sends poll request to a node and receives one or more poll response messages, depending on size message and needed fragmentation.
Comparison results are represented on figure 8 corresponding to a network with 5 nodes sending the same amount of data.
0 20 40 60 80 100 120 140 160
0 20 40 60 80 100 120 140 160 180
Data to be transmitted per stationDataresp (bytes)
Pollingcycle duration(ms) - -GTS
— CSMA/CA slotted
—CSMA/CA unslotted
(a) unacknowledged
0 20 40 60 80 100 120 140 160
0 20 40 60 80 100 120 140 160 180
Data to be transmitted per stationDataresp (bytes) Pollingcycle duration(ms)
(b) acknowledged
Figure 8. Access methods comparison with a 5 stations polling cycle
We observe that CSMA/CA unslotted provides lower cycle duration than GTS (except on very specific cases). But in what concerns CSMA/CA slotted, results are quite similar. GTS may provide lower cycle duration for low and average size messages whereas CSMA/CA is interesting for long messages, greater than 1 packet.
6. RELATIVE WORKS
Our results focuses on a specific mechanism pro- vided by IEEE 802.15.4 standard. Those results may be completed by (Koubâaet al., 2006b; Mišić and Fung, 2007) who propose simulation results concerning CSMA/CA slotted.
Use of wireless networks for industrial applica- tions may rely on the use of other standard, like IEEE 802.11 or IEEE 802.15.1.
IEEE 802.15.1 was developed in order to unbind equipments from their network wire. Due to some power limitation, its range of action is quite lim- ited (mostly around 10 m). Main applications of that protocol concern computer peripheral devices (printer, keyboard, cell phone, . . . ). Transmission of real-time traffic over a Bluetooth network has been studied under industrial cases(Lo Bello et al., 2005). Although that protocol is suitable for short range real-time communications, its intri- cate definition and high energy consumption mat- ter with its use with small mobile devices.
IEEE 802.11, first of them, is currently most used for computer wireless networking. It relies on a CSMA/CA medium access which is not deterministic. A second medium access, named PCF (Point Coordination Function), exists and it allows to assure the transmission of a time constrained traffic. Some publications (Bianchi,
2000; Krommenacker and Lecuire, 2005) focus on IEEE 802.11 performances to support industrial constraints.
7. CONCLUSION
That paper evaluates IEEE 802.15.4 temporal performances aimed to be used by industrial ap- plications. It underlines some heavy limitations concerning cycle duration and number of devices that can communicate within Contention Free Pe- riod.
It reveals some difficulties to use IEEE 802.15.4 for heavy time constrained applications. Moreover main commercial products based on that stan- dard are Zigbee ones. Zigbee protocol add more restrictions and it appears very rough to use such a technology for control command applications.
Use of IEEE 802.15.4 for more responsive use than supervision needs at least reconsideration of some standard values. It may also be possible to consider adaptation of some standard rules.
REFERENCES
Bianchi, Giuseppe (2000). Performance analysis of the IEEE 802.11 distributed coordination function.IEEE JSAC18, 535–547.
DYNAMITE — Dynamic Decisions in Mainte- nance(2005–2009).
Gutiérrez, José A., Edgar H. Jr Callaway and Ray- mond L. Jr Barrett (2003). LR-WPAN: En- abling Wireless Sensors with IEEE 802.15.4.
IEEE Computer Society (2006). Std 802.15.4.
Koubâa, Anis, Mário Alves and Eduardo Tovar (2006a). Energy and delay trade-off of the GTS allocation mechanism in IEEE 802.15.4 for wireless sensor networks. In:Wiley JCS.
Koubâa, Anis, Mário Alves, Bilel Nefzi and Ye- Qiong Song (2006b). Improving the IEEE 802.15.4 Slotted CSMA/CA MAC for Time- Critical Events in Wireless Sensor Networks.
Krommenacker, Nicolas and Vincent Lecuire (2005). Building Industrial Communication Systems based on IEEE 802.11g wireless tech- nology. In:ETFA. Vol. 1. pp. 71–78.
Lo Bello, Lucia, Mario Collotta, Orazio Mirabella and Thomas Nolte (2005). Approaches to support real-time traffic over bluetooth net- works. In:RTN. pp. 47–50.
Mišić, Jelena and Carol J. Fung (2007). The impact of master-slave bridge access mode on the performance of multi-cluster 802.15.4 network.Computer Networks51, 2411–2449.
Ramamurthy, H., B.S. Prabhu, R. Gadh and A.M.
Madni (2007). Wireless industrial monitoring and control using a smart sensor platform. In:
Sensors Journal. Vol. 7. pp. 611–618. IEEE.
